humanizer-tw

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [No Code] (SAFE): All provided files are Markdown-based documentation or reference materials. No executable scripts (Python, Node.js, or Shell) or tool logic are included in the analyzed set.\n- [External Downloads] (SAFE): The documentation refers to installation via GitHub or npx, which is standard for Claude Code skills. No suspicious runtime downloads or external package dependencies were detected.\n- [Indirect Prompt Injection] (LOW): The skill's primary function involves processing untrusted user text for transformation, which is an inherent surface for indirect prompt injection. However, this is expected behavior for a humanizer tool.\n
  • Ingestion points: User-provided text strings passed to the /humanizer-tw command.\n
  • Boundary markers: None are explicitly defined in the provided documentation files.\n
  • Capability inventory: CLAUDE.md specifies that the agent may use tools such as Read, Write, and Edit.\n
  • Sanitization: No input sanitization or validation logic is present in the provided reference files.\n- [Prompt Injection] (SAFE): No instructions designed to override agent safety protocols or reveal system prompts were identified.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:38 PM